[FFmpeg-devel] [PATCH 12/12] lavf/framecrcenc: do not hash side data

[James Almer]

On 5/14/2021 5:28 PM, Anton Khirnov wrote:
> Quoting James Almer (2021-05-14 14:42:10)
>> On 5/14/2021 5:01 AM, Anton Khirnov wrote:
>>> Quoting Michael Niedermayer (2021-05-10 16:06:01)
>>>> On Sun, Apr 25, 2021 at 09:03:20AM +0200, Anton Khirnov wrote:
>>>>> There are no guarantees that all side data types have the same
>>>>> representation on all platforms.
>>>>
>>>>> @@ -65,63 +51,6 @@ static int framecrc_write_packet(struct AVFormatContext *s, AVPacket *pkt)
>>>>>                 pkt->stream_index, pkt->dts, pkt->pts, pkt->duration, pkt->size, crc);
>>>>>        if (pkt->flags != AV_PKT_FLAG_KEY)
>>>>>            av_strlcatf(buf, sizeof(buf), ", F=0x%0X", pkt->flags);
>>>>> -    if (pkt->side_data_elems) {
>>>>> -        int i;
>>>>> -        av_strlcatf(buf, sizeof(buf), ", S=%d", pkt->side_data_elems);
>>>>
>>>> The number and type of the side data elements should not be affected
>>>> by the problem described.
>>>> Maybe we could leave that. Bugs could manifest as the absence or addition
>>>> of side data, seeing that in framecrc_write_packet() output may be usefull
>>>
>>> No strong opinion here - patches welcome I guess.
>>
>> I can do it, but it will be a "breaking" change, assuming there are
>> parsers that read the output of this muxer.
> 
> Imo frame* muxers are for testing purposes only and should not be
> considered stable.

Agree for framecrc as i mentioned in my other email, but framehash is 
versioned, so that one is stable.

Will send a patch to re-add side data amount of sizes to framecrc, then.